Local area educational inequality and self-rated health: A new approach to the Wilkinson hypothesis

Background: Wilkinson's inequality hypothesis states that an individual's health is affected not only by their own income, but also by the level of income inequality in their area. New insight on the hypothesis may be generated by considering alternative ways of operationalizing inequality.

Methods: Secondary analysis of Argentina's 2005 National Risk Factor Survey (N = 41,392). Multilevel binary logistic regression was used to examine how individual-level predictors (sex, age, education and household income) and local-area level predictors (educational inequality, measured by Gini coefficients of years of education, and mean educational attainment) affect the risk of poor self-rated health. Census data was used to derive local area measures.

Results: Individual-level measures of socio-economic status produce the expected gradients. Respondents in low and medium income groups are more likely to self-report being in poor health than respondents in high income groups (OR = 2.64, 95% CI = 2.38 – 2.94 and OR = 1.61, 95% CI = 1.45-1.79, respectively). Area-level educational inequality also generates a gradient. Respondents who live in the most unequal areas are most likely to report being in poor health (OR = 1.48, 95% CI = 1.26 – 1.74), followed by respondents in medium-high (OR = 1.40, 95% 1.18 – 1.67) and medium-low (OR = 1.19, 95% = 0.98 – 1.44) inequality areas, compared to low inequality areas, after adjusting for individual-level factors. Both individual- and area-level gradients remain stable after adjustment for an area's mean educational attainment.

Conclusion: Measuring inequality with years of education is a viable alternative to established income-based approaches.